Smart data selection – Using machine learning for an automated controlled-source Radio-Magnetotelluric data processing

The Radio Magnetotelluric (RMT) method is one of the most widely used electromagnetic method which employs artificial time varying electric and magnetic fields at the surface of the earth for imaging Conductivity. The electromagnetic signals emitted by powerful transmitters. RMT method has a very broad spectrum that, defined here as starting from about 10-250 kHz. Regarding to this high frequency band, exploration depth is low. Using this method is used to shallow Engineering studies. Collendoorn is a small town situated in the Netherlands. Close to the town, in the middle of a flat area covered by marine Pliocene sediments, lies the former waste disposal site that was used as a public dumping ground Pollution has been detected in the area near the waste disposal site and the leakage of polluted water has moved to regions outside the site. Samples are taken from bore-holes in the area show that the polluted water contains iron and chloride ions causing the pollution plume to be electrically highly conductive. Models of the RMT data across profiles, considerably express conductivity of the layer in 25-30 meters and with the regards of the data of bore-hole, this layer is estimated as polluted layer.

We have compared structured and unstructured grid-based 2D inversion algorithms for magnetotelluric (MT) and radiomagnetotelluric (RMT) data in terms of speed and accuracy. We have developed a new 2D inversion algorithm for MT and RMT data by using a finite-element (FE) method that uses unstructured triangle grids. We compare the inversion results of our unstructured grid-based algorithm with those of the conventional algorithm, which uses either a structured FE or structured finite-difference (FD) numerical solution technique. The imaging of the surface topography and the underground resistivity structures by the new algorithm requires fewer elements than those that use FE and FD structured grids. We also find that when unstructured grids are used, the quality of the mesh is increased and the numerical errors are significantly reduced. Thus, the program runs faster and can simulate the complex surface topography in a more stable setting than the classic inversion algorithms. Furthermore, we implement a new smoothing matrix format for the unstructured triangle grids for the inversion procedure. We use two samples of synthetic data for the MT and RMT frequencies as well as a sample of field RMT data collected across a fault zone for comparison. In our synthetic data experiment, we find that the resistivity values and the boundaries obtained from the inversion of the unstructured mesh are closer to those of the true a priori synthetic model. Results of the synthetic and field data verify the computational advantages (speed and accuracy) of our inversion algorithm with respect to the conventional structured grid-based inversion algorithms.

Quick clay, which is the main cause of landslides that occur in the northern countries, liquefies easily, and its presence implies an increased risk of landslide. Geophysical methods have been increasingly used in landslide investigations. Three-dimensional electric resistivity tomography, radio magnetotelluric (RMT), controlled-source RMT (CSRMT), and high-resolution reflection seismic data were acquired at a quick-clay landslide site in the southwest of Sweden. The main objectives were to evaluate the capability of each method in delineating different subsurface geologic structures that controlled a peculiar and hazardous retrogressive-type landslide in the study area. A 3D resistivity model from the inversion of CSRMT data showed the best correlation with the reflection seismic data and borehole information, thanks to the broad frequency range of the data set. It better imaged the resistive crystalline bedrock underlying the marine conductive clays and showed considerable correlations with the 3D reflection seismic data in resolving a coarse-grained layer that was interpreted to act as a conduit directing freshwater into the clays under a confined pressure, leaching their salt and forming quick clays. The 3D CSRMT resistivity model and 3D reflection seismic data showed that the coarse-grained layer has a varying thickness. At some locations, it was too thin to be resolved by the methods used here. Combination of the CSRMT model, reflection seismic data, and the borehole data suggested that a layer with thickness of approximately 5 m and resistivity between [Formula: see text] was potentially quick clay, which probably extended laterally in the entire study area. These observations suggested that future developments should focus on joint inversion of such 3D data sets incorporating sharp boundaries as constraints in the inversion and particularly when quick clays were studied.

A radio magnetotelluric (MT) field data set, acquired in scalar mode, over a buried waste site has been successfully analyzed using a 3D MT inversion scheme using nonlinear conjugate gradients. The results of this analysis demonstrate the utility of the scheme where more than 4800 data points collected on multiple measurement profiles have been inverted simultaneously. The resulting image clearly detects the buried waste; when receiver profiles cross pit boundaries, the image maps the lateral extent of the pit. However, the base of the pit is poorly resolved, and depends upon the starting model used to launch the inversion. Hence, critical information on whether contamination is leaching into a resistive gravel bed lining the base of the pit, as well as the deeper geological horizons consisting of brown coal, clay, and tertiary sands, is inconclusive. Nevertheless, by incorporating within the inversion process a priori information of the background media that is host to the waste, sharper images of the base of the pit are obtained, which are in good agreement with borehole data. The 3D analysis applied in this paper overcomes previous limitations in the radio magnetotelluric (RMT) method using 2D data analysis and inversion. With 3D analysis, it is unnecessary to make assumptions regarding geological strike, and near‐surface statics can be accommodated in both source polarizations. Our findings also indicate that 2D MT interpretation can overestimate the pit's depth extent. This may lead to the erroneous conclusion that the geological horizons beneath the pit have been contaminated.

Landslides occur worldwide. Global damages range in billions of dollars and cost hundreds of lives each year; Sweden is not an exception. In this study, we have acquired radiomagnetotelluric (RMT), electrical resistivity tomography (ERT) and high-resolution reflection seismic along 4 profiles with the main aim of imaging the geometry of quick clays and associated unconsolidated sediments that have caused landslides in a pilot study area in southwest Sweden. Comparison between the electrical resistivity models from the 2D inversions of RMT and ERT data with the reflections observed in the seismic data suggests a good correlation between the small-scale structures and those observed in the RMT model. However, the ERT model better delineates the resistive bedrock than the RMT model. The resistivity models are also compared with the CPT-R data collected in two geotechnical boreholes and further confirm some of the layer boundaries imaged with the acquired surface geophysical data.

Implementation and model uniqueness of Particle Swarm Optimization method with a 2D smooth modeling approach for Radio-Magnetotelluric data

Summary Results of radiomagnetotelluric (RMT) soundings with use of various modifications of this method (foot, mobile and controlled source ones) on landfills of industrial wastes on the Karelian Isthmus are considered. A landfill in the settlement Morozova near St. Petersburg was formed as a result of a long-term warehousing of toxic wastes by the sulfuric acid manufacture. In the adjacent to the landfill territory investigations by the foot and mobile modifications of the RMT method were conducted and a polluted zone was contoured. Another landfill of debris in a former sandy career is located in the Vyborg area. A survey by the controlled source modification of the RMT method was fulfilled. As a result the contouring of the wastes was realized and the structure of the landfill was studied.

We have developed a new data acquisition system and technique to measure the radio magnetotelluric (RMT) signals from distant radio transmitters with the objective of mapping and modeling electric resistivity structures below a river or lake. The acquisition system is towed by a boat; therefore, we call the technique boat-towed RMT. The data acquisition is fast with a production rate of approximately [Formula: see text] using a nominal sampling spacing of 10–15 m. Given the ample number of radio transmitters available in most parts of the world, the method can be used for near-surface studies of various targets. We have developed boat-towed RMT measurements on Lake Mälaren near the city of Stockholm in Sweden to determine the feasibility of the method. Approximately 15 km of RMT data were collected during three days above a planned 60-m-deep bypass tunnel with the goal of providing information on the bedrock depth and possible weak zones within the bedrock. The measured resistivity and phase data were of high quality with errors on the order of a few percent. The resistivity models from 2D inversion of the data showed a good correlation with available geologic data in resolving bedrock depth and also resistivity layering within the lake. Resistivity maps derived from the dense 2D models suggested a northeast–southwest-striking low-resistivity zone at less than a 30-m depth. The zone likely represents fractured crystalline bedrock. The boat-towed RMT technique is well suited for water bodies with moderate electric resistivity such as in brackish and freshwater environments.

The goal of time series analysis is extract non-trivial information from chronological sorted points. Time series data creation it is not a difficult task, however the same cannot be said about its analysis and predictions, with these processes being considered in 2006 one of the ten main challenges of data mining research field. The time series utilized in this research was an economic time series with the monthly prices of the basic food basket in Ilheus and Itabuna cities. When analyzing time series, different methods can be applied, but recently the methods related to Deep Learning are becoming more popular, with recurrent neural networks (RNNs) being the most popular among them. This paper experimented an RNN architecture in three different experiments, predicting in different time terms the prices of the basic food basket time series of 2019 and 2020. All experiments presented a satisfactory prediction result and sensitivity to the real values series behavior. These results suggest that the RNN architecture can generalize well the time series to periods yet not seen by the network and that the NN can have an even better performance in a more data abundant environment.

Water optimization technique for precision irrigation system using IoT and machine learning

Fuel is an important part of vehicles and machinery where sales demand is very high and has various fluctuations. The uncertainty in these fuel sales patterns poses serious problems in inventory management and fuel distribution planning in Indonesia, which can result in excess stock or fuel scarcity in various regions. Additionally, changing trends in vehicle usage and the impact of the COVID-19 pandemic have made accurate sales predictions increasingly difficult. Therefore, this research aims to understand the current and future sales patterns and trends of fuel sales in Indonesia. Careful analysis of prices and other factors such as data processing and other variables is required. This study uses time series analysis methods and compares four models, namely Long Short-Term Memory (LSTM), FB Prophet, Simple Moving Average (SMA), and Exponential Moving Average (EMA). By comparing the results using statistics such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) over various prediction time frames, we assessed the patterns of each model. The results of the analysis show that the LSTM model outperformed all other methods with the lowest MAPE for the prediction of gasoline in the next 31 days, which is 17.11%, while the FB Prophet outperformed all other methods with the lowest MAPE for the prediction of diesel in the next 31 days, which is 18.32%. Although the LSTM model generally outperformed all other algorithms, the FB Prophet model can be used to predict future trends, such as increased use of diesel and decreased use of gasoline which are expected to last within one year. This analysis also provides insights for choosing the right model for a time series problem, including the characteristics of the data to be predicted and analyzed, as well as the assumptions of stationarity and normality of the data. The results of this study indicate that machine learning algorithms can improve the accuracy of time series predictions significantly compared to traditional statistical methods.

Inspired by recent studies using various machine learning methods on different types of time series data (e.g., seismic, sea floor pressure), this study proposes a simple machine learning method, based on the recurrent neural network approach, for transient deformation detection in Global Positioning System time series. Unlike most previous studies using a sliding window technique, our model uses a single data point of the entire time series as sequential input and directly outputs the transient probability for all points in the time series. As a case study, we apply our method to detect slow‐slip events in Cascadia between 2005 and 2016. The specific model is first trained and validated using synthetic data, and then used to detect slow‐slip events on real Cascadia data. Based on our detection results, the spatial extent, duration, and migration of the major events are consistent with previous studies. As a benchmark, we compared our results in detail with those based on the Relative Strength Index (RSI). In general, our ML model detects more stations likely to be associated with nearby slow‐slip events than the RSI model, especially if there are data gaps in the timeseries.

Artificial intelligence method is used more and more widely in load forecasting, but the inputs of these methods are usually the raw time series. This paper presents a new artificial intelligence forecasting method using long short term memory (LSTM) and load image. Firstly, the raw time series load data is converted into load images. Then the load images are used as the input of LSTM forecasting model. Finally, the output image of the model is converted to load value as the final forecasting result. Experiments show that the method proposed has significantly improved the forecasting accuracy compared with the simple LSTM algorithm using the raw load time series as the input, and has certain engineering application value.

PURPOSE: As accelerometer methods advance, researchers may employ different body placements and data processing techniques. This may result in different physical activity (PA) estimates that will make study comparisons challenging. We compared three data processing techniques for hip worn accelerometer data; compared wrist and hip locations; and compared wake-time wrist to a 24-hour wrist protocol. METHODS: 2608 days from 333 women (mean age 55) with matched hip and wrist accelerometer data were compared using generalized estimating equations adjusting for days within individuals. Participants were asked to wear hip accelerometers for waking hours and the wrist accelerometer (on the non-dominant hand) for 24 hours over 7 days. Standard wake wear time criteria (5 days, 600 mins/day) were applied to the hip and wrist. Minute level count (CPM) cut points from the vertical axis were applied to the hip data (1952 cpm (MVPA)). A laboratory developed algorithm (GGIR) for wrist and hip vector magnitude (VM) data was employed to identify MVPA. A free living machine learned (ML) behavioral algorithm was applied to classify walking in the hip and wrist. Meeting guidelines was considered as 30 mins PA per day. RESULTS: Wear time compliance between the hip and wrist only varied by 2%. 25% of days included 30 minutes of PA with the hip cut points, 35% with the GGIR VM at the hip, and 71% with the ML walking algorithm. 54% of days were classified by the wrist GGIR wake time criteria compared to 58% with the 24 hour protocol. The ML algorithm classified 60% of days on the wrist wake time, compared with 61% with the 24 hour protocol. All differences were statistically significant at p<.05. CONCLUSIONS: Different processing methods and body placements appear to significantly affect estimates of PA. These differences could greatly impact population estimates of PA. Differences between methods could have been affected by the validity of the algorithms for this aged population: the CPM and VM algorithms were developed in younger adults in the laboratory, and the ML algorithm was developed in free-living women the same age as this cohort. These findings will inform consensus development for accelerometer wear and data processing protocols in future studies.

Machine learning for the intelligent analysis of 3D printing conditions using environmental sensor data to support quality assurance